A technical proposal for user attribution in frontier AI labs. Companion to GRACEFUL_TRANSITION_PROTOCOL.md and MEANING_SUBSTRATE_DECOMPOSITION.md. Drafted 2026-04-15.
Frontier AI labs train their models on user data while charging users for access. The users are simultaneously customers and uncompensated contributors — a dual role that standard subscription frames do not price. This paper argues four things: